Appendix: Methods for the Classification of Data from Open-Ended Questions in Surveys
Disputation 16 April 2024
Camille Landesvatter
University of Mannheim
Appendix Intro
Overview of Methods for Studies 1-3
Table 1: Overview Classification Methods and Examples. Source: Own depiction.
Characteristics of Open-Ended Survey Answers
Figure 1: The previous question was: ‘How often can you trust the federal government in Washington to do what is right?’. Your answer was: ‘[Always; Most of the time; About half of the time; Some of the time; Never; Don’t Know]’. In your own words, please explain why you selected this answer.
Appendix Study 1
Study 1: Results
Figure 1: Distribution of associations with known people across trust measures. Note: CIs are 95%, n=7,497.
Study 1: Results
Figure 2: Distribution of associations and their sentiment across trust measures. Note: CIs are 95%.
Study 1: Results
Figure 3: Associations and trust scores across different measures. Note: CIs are 95% and 90%.
Appendix Study 2
Study 2: Item Nonresponse by Condition
Figure 5: Item response rates by experimental condition and item.
Study 2: Sample Composition (Sample ↔︎ Population)
Table 1: Sample Characteristics in Comparison to US Census Data (2015).
Study 2: Covariate Balance (Text and voice)
Table 2: Covariate Balance between Text and Voice Condition.
Study 2: Results
Figure 6: Information Content Measures across Questions. Note. CIs are 95%, n_vote-choice: 830 (audio: 225, text: 605), n_future-children: 1,337 (audio: 389, text: 748)
Study 2: Results
Figure 6: Exemplary Survey Answers by Entropy.
Appendix Study 3
Study 3: Results
Figure 7: Sentiment Classification with three categories by classifier (BERT vs. GPT). Note. n=491 open-ended answers.
Study 3: Results
Figure 8: Linear model of sentiment and a five-category trust score (bi- and multivariate). Note. n=491 open-ended answers, GPT classification.
Study 3: Results
Figure 9: Sentiment Classification with five categories by classifier (AFINN vs. VADER). Note. n=496 open-ended answers.